DeepMind x UCL | Deep Learning Lectures | 8/12 | Attention and Memory in Deep Learning

Показать описание

Attention and memory have emerged as two vital new components of deep learning over the last few years. This lecture by DeepMind Research Scientist Alex Graves covers a broad range of contemporary attention mechanisms, including the implicit attention present in any deep network, as well as both discrete and differentiable variants of explicit attention. It then discusses networks with external memory and explains how attention provides them with selective recall. It briefly reviews transformers, a particularly successful type of attention network, and lastly looks at variable computation time, which can be seen as a form of 'attention by concentration'.

Download the slides here:

Find out more about how DeepMind increases access to science here:

Speaker Bio:

Alex Graves completed a BSc in Theoretical Physics at the University of Edinburgh, Part III Maths at the University of Cambridge and a PhD in artificial intelligence at IDSIA with Jürgen Schmidhuber, followed by postdocs at the Technical University of Munich and with Geoff Hinton at the University of Toronto. He is now a research scientist at DeepMind. His contributions include the Connectionist Temporal Classification algorithm for sequence labelling (widely used for commercial speech and handwriting recognition), stochastic gradient variational inference, the Neural Turing Machine / Differentiable Neural Computer architectures, and the A2C algorithm for reinforcement learning.

About the lecture series:

The Deep Learning Lecture Series is a collaboration between DeepMind and the UCL Centre for Artificial Intelligence. Over the past decade, Deep Learning has evolved as the leading artificial intelligence paradigm providing us with the ability to learn complex functions from raw data at unprecedented accuracy and scale. Deep Learning has been applied to problems in object recognition, speech recognition, speech synthesis, forecasting, scientific computing, control and many more. The resulting applications are touching all of our lives in areas such as healthcare and medical research, human-computer interaction, communication, transport, conservation, manufacturing and many other fields of human endeavour. In recognition of this huge impact, the 2019 Turing Award, the highest honour in computing, was awarded to pioneers of Deep Learning.

In this lecture series, research scientists from leading AI research lab, DeepMind, deliver 12 lectures on an exciting selection of topics in Deep Learning, ranging from the fundamentals of training neural networks via advanced ideas around memory, attention, and generative modelling to the important topic of responsible innovation.

Рекомендации по теме

Комментарии

*DeepMind x UCL | Deep Learning Lectures | 8/12 | Attention and Memory in Deep Learning*
*My takeaways:*
*1. Introduction **1:24*
1.1 Attention, memory and cognition 1:28
1.2 Attention in neural networks 2:50
-Implicit
-Can be checked through Jacobian
1.3 Explicit attention: hard attention, non-differentiable 17:00
-It has several advantages over implicit attention
--Computational efficiency
--Scalability (e.g. fixed size glimpse for any size image)
--Sequential processing of static data (e.g. moving gaze)
--Easier to interpret
-Neural attention models 19:24
-Glimpse distribution 20:25
-Attention with reinforcement learning 21:12
-Complex glimpse 22:46
*2. Explicit attention: soft attention, differentiable **26:27*
2.1 Basic 28:15
2.2 Attention weights 29:22
2.3 An example: handwriting synthesis with RNNs 32:40
2.4 Associative attention 38:38
2.5 Differentiable visual attention 45:30
*3. Introspective attention **49:23*
3.1 Neural Turing Machine 51:02
3.2 Selective attention 52:53
3.3 Content-based and location-based attention 55:28
3.4 Differentiable Neural Computer 1:12:04
*4. Further topics **1:13:51*
4.1 Self-attention in Transformers 1:14:00
*5. Summary **1:34:14*

leixun

From comment of Lei Xun (I added a 0.00 timestamp for the see chapters in the video)
0. Opening 0:00
1. Introduction 1:24
1.1 Attention, memory and cognition 1:28
1.2 Attention in neural networks 2:50
-Implicit
-Can be checked through Jacobian
1.3 Explicit attention: hard attention, non-differentiable 17:00
-It has several advantages over implicit attention
--Computational efficiency
--Scalability (e.g. fixed size glimpse for any size image)
--Sequential processing of static data (e.g. moving gaze)
--Easier to interpret
-Neural attention models 19:24
-Glimpse distribution 20:25
-Attention with reinforcement learning 21:12
-Complex glimpse 22:46
2. Explicit attention: soft attention, differentiable 26:27
2.1 Basic 28:15
2.2 Attention weights 29:22
2.3 An example: handwriting synthesis with RNNs 32:40
2.4 Associative attention 38:38
2.5 Differentiable visual attention 45:30
3. Introspective attention 49:23
3.1 Neural Turing Machine 51:02
3.2 Selective attention 52:53
3.3 Content-based and location-based attention 55:28
3.4 Differentiable Neural Computer 1:12:04
4. Further topics 1:13:51
4.1 Self-attention in Transformers 1:14:00
5. Summary 1:34:14

menesun

Alex Graves invented CTC and RNNT, which is basically a modern e2e ASR model in 2013. It created tens of thousands of research jobs, and he left to seek his desire. His journey is inspiring. He doesn't seek fame or money or status. He seeks the answer to his internal curiosity. I wanna live like him.

kimchi_taco

A no-nonsense detailed attention based lectures. A very well prepared lecture for all (beginners and experienced deep learning practitioner). Greatly recommended for all who want a context on how attention is first thought through in the research world. Thank you Alex. Enjoyed the lecture.

drpchankh

Great lecture! I really appreciate how he explains the thought process behind the new ideas.

barisdenizsaglam

Great Lecture ! Highly recommend anyone who is looking for indepth understanding of attention and different families of attention mechanism please watch this video . Its the best attention explanation available on the entire web.

stephennfernandes

This is good. So well explained. It's like a Neuralink knowledge upload to my brain. Thanks, Alex!

pw

This is one of my all-time favorite lectures, thanks for making this available. DNCs are very interesting.

peterdavidfagan

I like the "Thank you very much for your attention" punch line at the end.

agamemnonc

Dear DeepMind, the link for the slides seems to be valid. Can anyone fix that?

kaymengjialyu

wonderful! Thanks for putting these lectures out!!

marcelomanteigas

You are really brilliant, sir. I am from your friend country Bangladesh 🇧🇩. Hope you will be more and more helpful

Letsfeelthenaturee

Great lecture and big thanks to DeepMind for sharing this great content.

lukn

Looking for a lecture on attention mechanism..and This was the best.

ansh

thank you for this lecture, learned a lot about attention

ProfessionalTycoons

Another high quality course from deepmind, thanks !

Naghaas

For anyone that watched this lecture and his lecture from two years ago, is the difference large enough for me to watch the one from two years ago? Thanks

siyn

hey! i really enjoyed the machine lecturing! BUT!!!! your name is graves but i dont see a scar and i played u quite a bit in aram and also no cigar and also no shotgun and also no collector in ur item list in the background! PROPS FOR THE BEARD!!!

PresidentGollumSmeag

Excellent, Added To My Research Library, Sharing Through TheTRUTH Network...

robertfoertsch

Thanks Alex for the cool lecture and research!

GrigorySapunov

DeepMind x UCL | Deep Learning Lectures | 8/12 | Attention and Memory in Deep Learning

DeepMind x UCL | Deep Learning Lectures | 1/12 | Intro to Machine Learning & AI

DeepMind x UCL | Deep Learning Lectures | 6/12 | Sequences and Recurrent Networks

DeepMind x UCL RL Lecture Series - Introduction to Reinforcement Learning [1/13]

DeepMind x UCL | Deep Learning Lectures | 8/12 | Attention and Memory in Deep Learning

DeepMind x UCL | Deep Learning Lectures | 11/12 | Modern Latent Variable Models

DeepMind x UCL | Deep Learning Lectures | 4/12 | Advanced Models for Computer Vision

DeepMind x UCL | Deep Learning Lectures | 10/12 | Unsupervised Representation Learning

DeepMind x UCL | Deep Learning Lectures | 9/12 | Generative Adversarial Networks

DeepMind x UCL | Deep Learning Lectures | 2/12 | Neural Networks Foundations

DeepMind x UCL | Deep Learning Lectures | 12/12 | Responsible Innovation

DeepMind x UCL | Deep Learning Lectures | 7/12 | Deep Learning for Natural Language Processing

DeepMind x UCL RL Lecture Series - Exploration & Control [2/13]

DeepMind x UCL | Deep Learning Lectures | 5/12 | Optimization for Machine Learning

DeepMind x UCL | Deep Learning Lectures | 3/12 | Convolutional Neural Networks for Image Recognition

DeepMind x UCL RL Lecture Series - Theoretical Fund. of Dynamic Programming Algorithms [4/13]

DeepMind x UCL RL Lecture Series - Policy-Gradient and Actor-Critic methods [9/13]

RL Course by David Silver - Lecture 1: Introduction to Reinforcement Learning

DeepMind x UCL RL Lecture Series - Model-free Control [6/13]

DeepMind x UCL RL Lecture Series - MDPs and Dynamic Programming [3/13]

DeepMind x UCL RL Lecture Series - Multi-step & Off Policy [11/13]

Reinforcement Learning 1: Introduction to Reinforcement Learning

DeepMind x UCL RL Lecture Series - Function Approximation [7/13]

DeepMind x UCL RL Lecture Series - Model-free Prediction [5/13]

Grid cells - Caswell Barry, UCL